I got tired of the anxiety of usage tracking on Claude and Codex. Decided to try and use what local compute power I had (Macbook Pro M3 36GB) to offload some of the work from the large models.
I already had coding models running locally through Ollama, so this weekend I built a small toolkit that lets Codex, Claude Code, or just me from the terminal delegate those kinds of bounded tasks to a local model:
I vibe-coded a tool that lets Codex and Claude hand the boring little coding tasks to a local Ollama model
Contributions, issues, problems are more than welcome!
Contributions, issues, problems are more than welcome!